top of page

The Demands of AI over RPA Part 2

During July the RPA and AI Meetup in was held in London. Hosted by Bloom Search this was the fourth of their quarterly events. I've previously posted regarding the panel discussion, but there wasn't time to include some of the interesting audience questions.

The panel was hosted by Leigh Varnham (Director & Head of the RPA Unit for the Public Sector, at Capgemini), and attended by Chris Gayner (Director of Labs at Symphony Ventures), Steve Algieri (Independent Consultant) and Jason Engelbrecht (Head of Technology and Chief Technology Officer at Thomson Reuters).

In Part One our take away was "RPA is not for Christmas," from the audience questions there was a clear theme that "Automation as a Service is the future."

Which RPA Vendors are going to be bought, maybe by Google, Oracle or SAP?

Views varied across the team though all agreed that a the current market capitalisation of the main vendors now makes them an expensive option to buy. There was also too much overlap between all of the top vendors to make any consolidation of vendors likely.

However, the discussion moved to the point of consumption, for example Microsoft Azure or Amazon AWS, or at the orchestration level, where some opportunities for greater alliances are possible. Big brands have already been taking this approach creating partnerships with RPA Vendors.

Thoughts on Lean Six Sigma and Optimising Processes.

On this topic there was a broad consensus of opinion that the methodology is not the most important bit, but getting started and engaging the teams in a collaborative effort is.

The view was Lean Six Sigma is a great methodology, but having the right people and starting earlier provides the best outcomes.

The question of whether to streamline first or automate first was raised and the panel view was that it can work both ways, it largely depends on the culture of the business. The key element is to avoid silo'ed working.

How do you assess the complexity of a process?

There were a range of views on complexity, all of which are relevant considerations. One suggestion was to take a risk-based approach to evaluating processes, and therefore don't start with a process that will have a direct effect upon your customer or payments.

Lean Six Sigma was flagged as an approach with tools that would help to identify and evaluate the complexity of processes. Another was to take a data first approach and map the complexity of decision making and data movement to evaluate the level of difficulty. This led to the advice that you should not be afraid to abandon the process if you begin to identify it is too complex.

Financial Services organisations are keen on audits and currently trust humans more than robots, this leads to the need to add additional rigour into the process which can be time consuming but necessary to improve the level of confidence.

What do you see happening in the next 5 years?

  • Multi-Language support is a must

  • Interfaces will get simpler, hiding further complxity

  • Thinking will become "automation-first," rather than expecting people are the solution to any new process.

  • Automation as a Services (AaaS) - the BPO Market has already been disrupted

  • The vocabulary will change from automation and we will hear more about work, workers and workforce automation.

  • Easier, Cheaper, Faster, Cloud

What is your vision of the future?

Unsurprisingly a vision of the future generated a wide range of discussions, education and healthcare being both singled out as being the greatest opportunities for advancement. The concerns around data are rarely a problem when you are ill, people happily provide their data with a view it may help them or someone else to be cured.

Alexa, Google and others will become another channel and further services will be consumed via that way. It will lead to greater integration of different services.

The human experience will be re-engineered, designers working on autonomous vehicles are now describing themselves as an experience engineer as the technical needs to operate a vehicle diminish.

There was some scepticism about how long this will all take, the example of the tube system where there is limited automation, or planned automation, despite the simplicity of the problem. The feeling that children who are entering the workforce now will see very similar roles to the previous generation. However their children will have jobs we do not recognise today.

How do you handle data,. security and the black-box of AI?

RPA will become RPA+ as it incorporates elements of machine learning. Tagging and labelling of data to be useful for AI though is still a challenge.

From a security perspective AWS is now hosting for the US Government. The feeling was that in the future we will rely on Cloud Security - Google, Amazon and Microsoft - as it is better than what could be achieved independently.

On the topic of the Black Box of AI, particularly the use of deep learning, there is already research being carried out to link decisions back to how the AI was trained. The view is emerging that if the approach to training can be demonstrated as robust then the outcome can be trusted. Feature engineering a machine model can use 10,000 columns of data, it is incredibly complex to reverse engineer these situations.

An example of an ICO visit was also recounted and while successfully explaining how the decision was made, it required a significant amount of effort to achieve and was not easy to do. Countering this situation is the suggestion that a person must always remain accountable for the machine decisions.


If you haven't been along to one of the London RPA and AI Meetups the I can highly recommend popping along to see for yourself.

If you are looking for impartial advice on how to start or scale your RPA and AI journey then please contact us at Wzard Innovation to learn more about how working alongside a partner can accelerate the pace of your journey.


bottom of page